Anomaly Detection for Aviation Safety Based on an Improved KPCA Algorithm

Xiaoyu Zhang,Quan Gan,Jiusheng Chen

doi:10.1155/2017/4890921

Abstract

Thousands of flights datasets should be analyzed per day for a moderate sized fleet; therefore, flight datasets are very large. In this paper, an improved kernel principal component analysis (KPCA) method is proposed to search for signatures of anomalies in flight datasets through the squared prediction error statistics, in which the number of principal components and the confidence for the confidence limit are automatically determined by OpenMP-based K-fold cross-validation algorithm and the parameter in the radial basis function (RBF) is optimized by GPU-based kernel learning method. Performed on Nvidia GeForce GTX 660, the computation of the proposed GPU-based RBF parameter is 112.9 times (average 82.6 times) faster than that of sequential CPU task execution. The OpenMP-based K-fold cross-validation process for training KPCA anomaly detection model becomes 2.4 times (average 1.5 times) faster than that of sequential CPU task execution. Experiments show that the proposed approach can effectively detect the anomalies with the accuracy of 93.57% and false positive alarm rate of 1.11%.

Highlights

Onboard Quick Access Recorder (QAR) and Flight Data Recorder (FDR) can record more than 600 parameters sampled at 1 Hz
An improved kernel principal component analysis (KPCA) solution is proposed for efficient anomaly detection
The radial basis function (RBF) parameter is optimized by GPU and OpenMP-based K-fold cross-validation is adopted for training KPCA anomaly detection model

Summary

Introduction

Onboard Quick Access Recorder (QAR) and Flight Data Recorder (FDR) can record more than 600 parameters sampled at 1 Hz. Kernel principal component analysis (KPCA) algorithm was used by Cho et al [5] for fault identification in process monitoring. There were two significant problems for anomaly detection in the KPCA-based method: computation performance and adaptability. For the KPCA-based method, the kernel matrix was defined for the principal component. Because the parameter in the kernel function, the number of principal components, and the confidence for the confidence limit should be set before anomaly detection by KPCA method, the adaptability of the KPCA method was largely limited. The method proposed in this paper is based on KPCA and aims at anomaly detection in the flight.

Kernel Principal Component Analysis

Improved KPCA Anomaly Detection Model

Numerical Experiments and Discussions

Findings

Conclusion